Vivek Tiwari
Machine Learning Engineer — GenAI, RAG, and MLOps
vivektiwari
San Francisco, California
Seattle, Washington
Joined December 2025
Network
461 connectionsCCPNCK
STCEIM
ABAAKK
CMGDPK
ABDKCA
NSIM
IMKT
Network
461 connectionsSummary
Practitioner of GenAI and production RAG systems — Vivek builds and ships Retrieval-Augmented Generation and agentic GenAI solutions aimed at reliable, production-grade performance. His portfolio and résumé note work on ingestion/ETL for unstructured medical and document text, vector search and Pinecone-based retrieval, LLMOps pipelines, and agent orchestration for real systems. twine+1
Open-source author and technical writer — he publishes medium/TowardsAI articles and maintains GitHub projects that demonstrate applied ML engineering (document intelligence, NdLinear experiments, RL projects). His writing combines implementation details, benchmarking, and reproducible code. medium+2
Focus on document intelligence and structure-aware modeling — several projects (DOCAI and related repositories) and articles emphasize NdLinear and structure-preserving layers for layout-aware transformers, indicating a specialty in form and document understanding. github+1
Experience in healthcare/clinical NLP workflows and productionization — past work at Docplus and IU shows emphasis on building inference APIs, chatbots/multimodal triage, and production ML pipelines with data quality and compliance considerations for medical contexts. wayup+1
Work
Education
Projects
Writing
Revolution in Neural Network Layers: How NdLinear Outperforms Traditional Linear Layers
April 1, 2025A technical article explaining NdLinear (N-dimensional linear layers), comparing nn.Linear vs NdLinear mathematically and empirically, and showing benefits for document intelligence and structured data.
From Pixels to Plans: Cracking 2048 with Reinforcement Learning and Beam Search
March 1, 2025A deep-dive project writeup comparing PPO-based RL agents and deterministic Beam Search for mastering the game 2048, including architecture, results, and code to reproduce experiments.